Azure AI Foundry Resource
Does ZopNight manage Azure AI Foundry Resource?
Azure AI Foundry resources are Cognitive Services accounts of kind AIServices, and the account is where token and PTU meters actually bill. ZopNight expands each account into its projects, model deployments, and connections, then attributes Cost Management spend across that hierarchy so AI cost lands on the team that generated it.
Rules that fire on Azure AI Foundry Resource
No active rule family targets Azure AI Foundry Resource today. Rules that used to are retired, and retired rules publish no pages and fire no findings. Scheduling and permissions coverage are unaffected.
At a glance
| Field | Value |
|---|---|
| Scheduling notes | discovery and topology only. |
Azure AI Foundry is Microsoft’s unified platform for building generative AI applications, exposing models, agents, and evaluation tooling. Foundry resources anchor the deployments and compute where AI spend concentrates.
Where Foundry spend actually accrues
A Foundry resource is a Cognitive Services account of kind AIServices, and the account is the billing anchor. Token consumption from pay-as-you-go model calls and the reserved capacity of provisioned deployments both land on the account’s meters; the child deployment rows themselves carry no charge of their own. In practice that means the invoice tells you which account spent, while only the expanded hierarchy underneath can tell you which model, project, or team drove the number.
Expanding the account into its children
Discovered via the AI enricher with child projects, deployments, and connections expanded beneath it. Cost Management billing attributes spend across the Foundry hierarchy. ZopNight inserts synthetic grouping nodes (Model Deployments, Agents, Content Filter Policies) between the account and each cluster of children, and keeps the account as the grandparent, so attribution stays intact even when an estate runs many Foundry accounts side by side.
Sprawl signals on Foundry accounts
The recurring failure mode is one account per experiment: a hackathon or proof of concept creates a fresh Foundry resource, deploys a model or two, and moves on. Each abandoned account may still hold provisioned deployments billing reserved capacity, connections into production data services, and agents nobody remembers configuring. A second pattern is the untracked account outside the platform team’s subscription, invisible until its meters show up in a monthly review. Inventorying every account and its children is what makes both patterns reviewable.
Reaching the account in the Azure portal
Azure portal → Azure AI Foundry lists the accounts, or filter Cognitive Services accounts to the AIServices kind. Day-to-day management of projects, deployments, and agents happens in the separate AI Foundry portal, which scopes its view to one account and project at a time; the Azure portal remains the place to see every account a subscription actually contains.